NCJRS Study of Megan's Law says it is useless

Principal Investigator:
Kristen Zgoba, Ph.D.
New Jersey Department of Corrections
Research & Evaluation Unit
Trenton, NJ 08625
(609) 777-4256
kristen.zgoba@doc.state.nj.us

Co-Principal Investigator:
Philip Witt, Ph.D.
Associates in Psychological Services, P.A.
25 N. Doughty Ave.
Somerville, NJ 08876
(908) 526-1177
phwitt@optonline.net

Research Assistant/ Data Manager:
Melissa Dalessandro, M.S.W.
New Jersey Department of Corrections
Research & Evaluation Unit
Trenton, NJ 08625
(609)-292-5670
melissa.dalessandro@doc.state.nj.us

This document is a research report submitted to the U.S. Department of Justice. This report has not been published by the Department.
Opinions or points of view expressed are those of the author(s) and do not necessarily reflect the official position or policies of the U.S. Department of Justice.

The author(s) shown above used Federal funds provided by the U.S. Department of Justice and prepared the following final report:

Document Title: Megan’s Law: Assessing the Practical and Monetary Efficacy

This report has not been published by the U.S. Department of Justice.
To provide better customer service, NCJRS has made this Federally-funded grant final report available electronically in addition to traditional paper copies.

necessarily reflect the official position or policies of the U.S. Department of Justice.

Some hyperlinks which were not in the original document have been added for those who wish to conduct further research.

This document has been posted in .htm format along with related documents, essays, and opinions on:

The research that follows concerns the various impacts of community notification and registration laws (Megan’s Law) in New Jersey. Although this report includes a variety of interesting findings and many ideas that will be explored upon post grant period, this research was embarked upon, in general, to investigate: 1) the effect of Megan’s Law on the overall rate of sexual offending over time; 2) its specific deterrence effect on re-offending, including the level of general and sexual offense recidivism, the nature of sexual re-offenses, and time to first re-arrest for sexual and non-sexual re-offenses (i.e., community tenure); and 3) the costs of implementation and annual expenditures of Megan’s Law. These three primary foci were investigated using three different methodologies and samples. Phase One was a 21-year (10 years prior and 10 years after implementation, and the year of implementation) trend study of sex offenses in each of New Jersey’s counties and of the state as a whole. In Phase Two, data on 550 sexual offenders released during the years 1990 to 2000 were collected, and outcomes of interest were analyzed. Finally, Phase Three collected implementation and ongoing costs of administering Megan’s Law. The following points highlight the major findings of the three phases of the study.

· New Jersey, as a whole, has experienced a consistent downward trend of sexual offense rates with a significant change in the trend in 1994.

· In all but two counties, sexual offense rates were highest prior to 1994 and were lowest after 1995.

· County trends exhibit substantial variation and do not reflect the statewide trend, suggesting that the statewide change point in 1994 is an artifact of aggregation.

· In the offender release sample, there is a consistent downward trend in re-arrests, reconvictions and re-incarcerations over time similar to that observed in the trend study, except in 1995 when all measures spiked to a high for that period. This resulted in

· significant differences between cohorts (i.e., those released prior to and after Megan’s Law was implemented).

· Re-arrests for violent crime (whether sexual or not) also declined steadily over the same period, resulting in a significant difference between cohorts (i.e., those released prior to and after Megan’s Law was implemented).

· Megan’s Law has no effect on community tenure (i.e., time to first re-arrest).

· Megan’s Law has no effect on the type of sexual re-offense or first time sexual offense (still largely child molestation/incest).

· Megan’s Law has no effect on reducing the number of victims involved in sexual offenses.

· Sentences received prior to Megan’s Law were nearly twice as long as those received after Megan’s Law was passed, but time served was approximately the same.

· Significantly fewer sexual offenders have been paroled after the implementation of Megan’s Law than before (this is largely due to changes in sentencing).

· Costs associated with the initial implementation as well as ongoing expenditures continue to grow over time. Start up costs totaled $555,565 and current costs (in 2007) totaled approximately 3.9 million dollars for the responding counties.

· Given the lack of demonstrated effect of Megan’s Law on sexual offenses, the growing costs may not be justifiable.

On July 29, 1994, Jesse Timmendequas, a sex offender who had been released after serving a maximum sentence in a New Jersey correctional facility, raped and murdered seven-year-old Megan Kanka in Hamilton, New Jersey. The intense community reaction that followed extended well beyond the state. One expression of community outrage was the enactment of laws to notify the public of the presence of sex offenders living and working in their community. The premise was, and still is, that with this knowledge, citizens will take protective measures against these nearby sex offenders. As Beck, Clingermayer, Ramsey and Travis (2004) note, “Exactly what action is expected is not clear, but it is hoped that, armed with this critical information, citizens will work on their own or in concert with government to make their neighborhoods safer” (p. 142).

During the following decade, all 50 states and the District of Columbia enacted some version of such community registration and notification laws, collectively referred to as “Megan’s Laws” (Presser & Gunnison, 1999; Zevitz & Farkas, 2000). Although a few states, such as Washington, had enacted community notification laws prior to 1994, the federalization of community notification laws in 1996 created strong incentives for other states to follow suit (Presser & Gunnison, 1999).

The legislation known as Megan’s Law, includes both registration and notification. Sex offenders must register their addresses with local police jurisdictions within a specified time of release from prison. By way of the registration process, the public is then notified of the offender’s presence in the neighborhood. The goal of notification is to inform both the public and past victims so that they can protect themselves accordingly. As with other states, registration and notification are separate steps in New Jersey, but are often referred to as one process. In New Jersey, offenders are placed into one of three tiers, representing a hierarchy of potential risk of an offender’s re-offense. A risk assessment instrument is used to predict the offender’s likelihood of re-offense, which ultimately determines placement into the tier. Tier one represents the lowest risk and requires only notification of law enforcement officials and the victims. Offenders are considered low risk and eligible for a tier one placement if they received a low risk assessment score and are on probation / parole, receiving therapy, employed and free of alcohol and drugs. A tier two classification represents a moderate risk of a re-offense. It requires notification of organizations, educational institutions, day care centers and summer camps. The factors for placement into a tier two category include a moderate to high risk assessment score, failure to comply with supervision, lack of employment, abuse of drugs or alcohol, denial of offenses, lack of remorse, history of loitering or stalking children and making threats (Brooks, 1996; Matson & Lieb, 1997; Witt & Barone, 2004). Tier three offenders are those who are predicted to present the greatest risk to re-offend. This category has generated the most legal resistance because it calls for the broadest level of notification. The entire community is notified through posters and pamphlets. The factors necessary for the placement into a tier three category are a high probability of re-offending evidenced by a particularly heinous instant offense or a high-risk assessment score, repetitive and compulsive behavior, sexual preference for children, failure or refusal of treatment, denial of the offense and lack of remorse (Brooks, 1996; Rudin, 1996; Witt & Barone, 2004).

Despite their existence for over a decade, little work has been done to examine the effectiveness of these laws on sexual offense rates. A few researchers, such as Beck and colleagues (2004), have conducted surveys to determine what protection methods community members use when given information regarding the presence of sex offenders. Beck and colleagues (2004) approach their research from the viewpoint that community notification laws were enacted more to change the behaviors of potential victims than those of potential sexual recidivists. In this study, Beck and colleagues (2004) differentiated between two types of protective measures: (1) “self-protective measures,” or behavioral measures initiated by the potential victims themselves; and (2) “altruistic protective measures,” or behavioral measures initiated by family members to protect other household members (e.g., their children) (Beck & Travis, 2002). These studies found that community notification did, in fact, increase altruistic behaviors by community members to protect members of their households, although the findings are inconsistent with regard to whether self-protective behaviors increased after community notification. Because of these results, Beck and colleagues (2004) posit that it is not the enactment of community notification laws themselves that influences protective behaviors, but the community members’ perceived risk of victimization (also measured in these surveys) that mediates these behaviors. This mediating factor presents problems for identifying the true effect of these laws on sexual recidivism rates.

A few studies have also surveyed sex offenders to determine the impact that community notification laws have had upon them. Tewksbury (2005) found that social stigmatization, loss of relationships, employment and housing, and both verbal and physical assaults were experienced by a significant minority of registered sex offenders (see also Tewksbury & Lees, 2006). Zevitz and Farkas (2000) also found that a majority of sex offenders reported negative consequences, such as exclusion from residences, threats and harassment, emotional harm to their family members, social exclusion by neighbors, and loss of employment. Furthermore, according to many tier three offenders interviewed, these laws would not deter them from committing future sex offenses (Zevitz and Farkas, 2000). In fact, Presser and Gunnison (1999) suggest that notification laws may be counterproductive in that public scrutiny causes additional stress to offenders who are transitioning back into the community. The fear of exposure may cause offenders to avoid treatment, and in the case of pedophiles, may encourage offenders to seek out children as a result of adult isolation. If these assumptions are true, the risk of recidivism may be increased (Presser & Gunnison, 1999), or at least such factors would work against any protective measures taken, thus lessening or eliminating any positive effect of the law.

None of the aforementioned research, however, addresses the critical question of whether community notification and registration laws actually reduce sex offense rates (primary offenses or re-offenses) in the communities in which the laws are applied, or what patterns of sexual offense rates appear. Despite Megan’s Laws being in effect in all 50 states, only one study was found that examines pre and post-Megan’s Law sex offense rates. That study, conducted in the state of Washington, compared sexual recidivism rates between two groups of sexual offenders: one released three years prior to the implementation of community notification laws in that state, and one released three years after the implementation. The pre and post- target groups were those most likely to be affected by the law (i.e., those who would qualify for tier three classification). To account for population differences, offenders in both groups were matched on the number of sex convictions and the type of victim (i.e., adult or child) (Schramm & Milloy, 1995). Their analysis of potential group differences revealed that at the end of 54 months (four-and one- half years “at risk”), there was no statistically significant difference in the arrest rates for sex offenses between the two groups (19 percent versus 22 percent). However, the study did find that notification had an effect on the time of the next arrest for any type of offense.

Offenders subject to notification were arrested for new crimes much more quickly than were offenders not subject to notification. (Schramm & Milloy, 1995).

These results suggest that Megan’s Laws may not be effective in reducing recidivism rates. One can make a case, in fact, that Megan’s Law, at least as implemented in Washington, had no effect on the rate of sex offense recidivism, although it may result in more rapid detection of new sex offenses (see discussion in Pawson, 2002).

This lack of outcome studies means that Megan’s Laws constitute an untested mandate in the domain of empirical research. Despite widespread community support for these laws, there is virtually no evidence to support their effectiveness in reducing either new first-time sex offenses (through protective measures or general deterrence) or sex re-offenses (through protective measures and specific deterrence).

The study described below investigates various impacts of community notification and registration laws (Megan’s Law) in New Jersey. The primary areas of study are: 1) the effect of Megan’s Law on the overall rate of sexual offending over time; 2) its specific deterrence effect on re-offending, including the level of general and sexual offense recidivism, the nature of sexual re-offenses, and time to first re-arrest for sexual and non-sexual re-offenses (i.e. community tenure); and 3) the costs of implementing and maintaining Megan’s Law. These three primary foci were investigated using three different methodologies and samples.

Phase One was a 21-year (10 years prior and 10 years after implementation, and the year of implementation) trend study of sex offenses in each of New Jersey’s counties and the state as a whole. In Phase Two, data on 550 sexual offenders released during the years 1990 to 2000 were collected, and outcomes of interest were analyzed. Finally, Phase Three collected implementation and ongoing costs of administering Megan’s Law.

This study attempts to remedy one aspect of the gap between the lack of research and the legislation, by examining the trend of sexual offense rates between and within the 21 counties of New Jersey from 1985 through 2005. The study was conducted in New Jersey, the state in which Megan Kanka was a victim and the subsequent origin of Megan’s Law. Phase One is a trend study, which will provide information on whether statistical differences exist in sex offending arrests before and after the implementation of Megan’s Law.

The trend analysis focuses on the pattern of sexual offense rates in New Jersey over a 21-year timeframe while comparing them to drug offense rates and non-sexually based offending rates. The data represent crime rates for the state as a whole and for each of the 21 counties for the ten years prior to the legislation and the ten years after the enactment of the legislation and includes the first full year in which Megan’s Law was implemented (i.e., 1995).

The purpose of this study is to determine whether Megan’s Law had an effect on the rate of sexual offending in New Jersey. Several different analyses were conducted to answer this primary question. First, a trend analysis of New Jersey sex offense rates pre and post-Megan’s Law implementation provides both a visual and statistical test of effectiveness. Second, aggregation sometimes masks important differences at a lower level. Therefore, the same trend analyses were conducted on each of the 21 counties in New Jersey. Third, historical effects broader than that solely for sex offenses may be responsible for observed changes (i.e., an observed effect of Megan’s Law may be spurious). Two comparative analyses at the state level were conducted to contrast changes in rates of sex offenses to other offenses (i.e., drug and other non-sex/non-drug) over the same period of time. These additional analyses were made in an effort to place sex crimes in the context of overall crime and a specific crime (drugs) that has been subjected to several types of legislation.

This study is based upon a simple pre-post research design to determine whether any significant changes in the rates of sexually based offenses reported by law enforcement agencies occurred after the implementation of New Jersey’s Megan’s Law in late 1994. Rates for sexually based offenses, non-sexually based offenses, and drug offenses were collected for the years 1985 through 2005 in order to construct a comparative trend analysis. Data for the three types of crime were collected for all 21 counties of New Jersey, using Uniform Crime Report (UCR) numbers for years 1985 through 2005.

Prevalence rates for the three offense categories were established using population estimates from the Department of Labor’s Bureau of Labor Statistics. The Department of Labor’s population estimates for New Jersey were cross-referenced with the Sourcebook of Criminal Justice Statistics, a yearly federal government publication. Because no significant differences in population estimates were found between these two sources, UCR numbers were used for trend analyses conducted in this study. In order to compare state and county trends in sexually based offenses, non-sexually based offenses, and drug abuse violations, UCR aggregate numbers and prevalence rates for years 1985-2005 were entered into an Excel spreadsheet and SPSS.

Uniform Crime Report statistics are based upon number of arrests, and as such, use of the term “offenses” in this study refers to number of reported arrests. Three crime categories were used for trend analysis comparisons: 1) sexually-based offenses, 2) drug offenses, and 3) other offenses (non-sex/non-drug). Analyzing the single set of sex offense rates for the 21-year time span provides an initial test of rate change. Across the US, crime rates in general have been dropping since the late 1990’s. The inclusion of all New Jersey non-sex/non-drug crime rate trends presents a visual contrast: (1) to confirm/disconfirm the national trends, and (2) to contextualize the sex offense rate trend within the general trends. Other offenses allow a control for New Jersey specific historical factors that might influence rates across crime categories, such as increased or decreased enforcement or prosecutorial budgets, the number of police or probation officers, or aggressiveness of prosecutors’ and the State Attorney General’s offices.

Drug offenses, like sex offenses, have been the target of law enforcement policies. Although drug offense rates may change over time based upon what drugs are most common, drug arrests rates are also particularly vulnerable to changes in federal and local policies and law enforcement efforts. Furthermore, although the contrast between drug and sex crimes may not be immediately obvious, the inclusion of drug offense rate trends provides an opportunity: (1) to demonstrate the variations in rates over time, and (2) to evaluate whether these variations have similar patterns to those of sex offense rates.^[1]

These crime categories were based on the state’s Uniform Crime Report (UCR), a yearly statistical report based upon crimes reported to law enforcement agencies throughout the State of New Jersey. Definitions of certain sexually-based crimes, such as rape, were clarified via phone interviews with the New Jersey State Police in December 2006. In addition, legal definitions of specific crimes (e.g. endangering the welfare of a child) were verified by reviewing Title 2C of the New Jersey Code of Criminal Justice in LexisNexis Academic. Because “Rape” is designated as a separate category by the UCR, the UCR categories “Rape” and “Sex Offenses” were combined under the category “Sexually-Based Offenses” for the purposes of this study. The category “Non-Sexually Based Offenses” is comprised of all UCR categories except “Rape”, “Sex Offenses” and “Drug Offenses”. Furthermore, “Drug Offenses” included various types of drug crimes, such as the manufacturing and distribution of controlled substances, possession with the intent to sell and distribution of a controlled substance within a school zone.

In studies of this type, typically a simple pre-post test of rates is conducted to determine whether an intervention is successful. Given that these data are points in time, namely crime rates by year, time based strategies are commonly used, including time-series/ARIMA models and regression discontinuity designs that allow for temporal autocorrelation. These analyses are constructed based upon a known change point. Although it is known that Megan’s Law was passed in late 1994, it is not known when the agencies charged with implementing the law were fully prepared to do so. Further, Megan’s Law may not have been uniformly implemented across the state at a standardized point. The earliest change point that might be attributed to the legislation, therefore, is between 1994 and 1995. Given delayed implementation, the true effect of the legislation may occur during a subsequent year. For this study, a method is required that will allow for the detection of such delayed effects.

Several authors have considered the problem of change-points (see Pettitt, 1979 for a brief review). Some make assumptions regarding the nature of the pre and post-change sample distributions. Most assume that the change-point is known. Pettitt (1979) offers a solution to the crime trend problem by suggesting a method of determining the most probable point of change and using a non-parametric procedure to test for significance. The logic of Pettitt’s argument is summarized below.

Assume a sequence of random variables; X1, X2, ... , XT and a change-point at , where Xt for t = 1, ... , have a distribution function of F1(x) and Xt for t = +1, ... , T have a distribution function of F2(x) and F1(x) F2(x). Since the change-point is unknown, T-1 two sample comparisons are necessary. In the complete sample of T,

where Wt is the sum of the ranks of all observations from 1 to t. This produces a U statistic for each point in the time series comparing the mean of the series prior to t with the mean of the series after t. A version of the Mann-Whitney U statistic, used to test that the two samples, X1, ... , Xt and Xt +1 , ... , XT , come from the same population, is applied to the maximum U value:

where the approximation holds accurate to two decimal places for p< .5 (Pettitt, 1979).

This analysis employs this technique used to determine significant differences when the change point is unknown. This technique was selected specifically because we did not want to make any assumptions regarding the implementation phase. In most cases, where a law requires changes to procedure, the effect is likely to be delayed by some unknown period.

Data from the 21 New Jersey counties were entered separately, the New Jersey total was aggregated from the counties’ summary numbers and the resulting rates were adjusted for year-to-year population changes at the state level. For each county and for the state as a whole, the yearly rates were rank ordered and a Mann-Whitney U test was performed to test for a change in trend. Thus, for the state and for each county, every year is tested as a potential change point.

The results are organized into two major sections. The first section presents the trend analysis for both the state and for the individual counties. The second section contrasts the sex offense trend rates to trends in other offenses (i.e., drug and other non-sex/non-drug) over the same period of time.

Figure 1 displays the rates of sex offenses for New Jersey as a whole from 1985 to 2005. The rates varied from 51 offenses in 1986 to a low of 29 offenses per 100,000 population in 2005. In general, there is a consistent downward trend.

Individual counties varied substantially both between counties and within counties over time. Table 1 presents summary statistics of each county and the state as a whole. Counties varied in population size from under 100,000 population in the smallest counties of Cape May, Salem, and Warren, to over three quarters of a million residents in the largest counties of Bergen and Essex. The population size of the county is not consistently related to the rate of sex offenses. For example, one of the largest counties, Essex County (Newark), has a relatively high rate of offenses (68), whereas the largest county, Bergen, has a relatively low rate (32). In contrast, the highest rate of offenses is in one of the smaller population counties, Cumberland. In the smallest counties, Cape May has a rate of 72 offenses per 100,000, whereas Warren has a rate of 36.

In 19 of the 21 counties, the year with the highest rate of sex offenses occurred before 1994; Passaic and Sussex Counties were the exceptions. In 19 of the 21 counties, the year with the lowest rate of sex offenses occurred after 1995; Morris and Passaic Counties were the exceptions in this case.

The rank trend tests (Mann-Whitney U tests) revealed that (1) six counties had no statistically significant change point (Bergen, Hunterdon, Mercer, Morris, Passaic and Sussex), and (2) an additional six counties had a change point that preceded Megan’s Law (Burlington, Camden, Monmouth, Salem, Somerset, and Union). This means that only nine counties have a change point after Megan’s Law was passed with the years of change falling between 1994 and 1998. One final observation of county contrast should be noted. In several cases, counties had substantial drops in sex offenses after Megan’s Law was enacted. However, in the last several years these counties have had substantial increases in sex offense rates (analyses not shown). This is true, for example, of Ocean, Hudson, and Warren Counties.

Table 1. County and State Summary Statistics for Sex Offenses

County	Population in 1994	Average Rate	Highest (Year)	Lowest (Year)	Change Year
Atlantic	236,589	71.0	128 (1991)	31 (2005)	1994
Bergen	848,392	32.0	71 (1988)	24 (2002)	n.s.
Burlington	407,060	30.4	51 (1985)	16 (2002)	1993 **
Camden	508,479	42.9	97 (1986)	29 (2005)	1989 **
Cape May	99,561	72.2	111 (1992)	38 (2003)	1995
Cumberland	144,544	91.1	127 (1992)	63 (2001)	1998
Essex	784,460	67.6	95 (1990)	35 (2004)	1994
Gloucester	242,161	35.6	53 (1993)	22 (2005)	1997
Hudson	572,720	44.0	56 (1993)	33 (2001)	1998
Hunterdon	113,522	18.3	32 (1985)	9 (2004)	n.s.
Mercer	335,229	46.4	67 (1986)	35 (2005)	n.s.
Middlesex	701,090	27.1	38 (1985)	19 (2004)	1998
Monmouth	577,069	37.6	56 (1988)	25 (2002)	1992 **
Morris	439,533	23.1	33 (1986)	15 (1993)	n.s.
Ocean	461,152	24.7	38 (1993)	15 (2001)	1996
Passaic	478,164	50.8	82 (1997)	36 (1988)	n.s.
Salem	64,691	50.8	78 (1991)	27 (2005)	1992 **
Somerset	262,243	18.9	27 (1988)	10 (1998)	1991 **
Sussex	137,021	21.8	32 (1999)	12 (2005)	n.s.
Union	504,864	30.0	53 (1986)	13 (2004)	1993 **
Warren	95,762	36.4	63 (1987)	14 (2001)	1996
NEW JERSEY		39.8	51 (1986)	29 (2005)	1994

** Change point precedes implementation point

Also, many counties demonstrated a predictable “jump” after Megan’s Law was implemented. After a large initial drop in rates, there was a large rebound in sexual offenses (but not as high as pre-Megan’s Law levels), followed by a continued decline. One example of this phenomenon is Cumberland County. As can be seen in Figure 2, the large dip at year 10 (1994) is followed by a spike the following year and then returns to a downward trend. This spike in sexual offenses most likely reflects increased surveillance and arrests, rather than increased offending.

Although individual counties vary, the aggregate state statistics indicate a significant change in trend in the year 1994 (MW-U=110.0, p<.001). Figure 3 displays the rates before and after the implementation of Megan’s Law. The upper line represents sex offenses for the years 1985-1994, and the lower line represents sex offenses for the years 1995-2005. Superimposed on the yearly rates is a linear trend line. There are two important differences between these trend lines. First, beginning in 1995 the rate of sex offenses never again approaches the pre-1994 levels (i.e., the intercept and average are different). Second, the slope is steeper in the post-Megan’s law period. This is particularly notable, since sex offenses are low base rate crime. The fact that the decrease accelerates as the number of crimes decreases is unexpected. In fact, one might expect that an effective intervention would exhibit diminishing returns over time. This is not the case in this instance.

The aggressiveness with which arrest, prosecution and surveillance of specific crimes is pursued changes over time. After Megan Kanka’s death at the hands of a convicted sex offender, public sentiment demanded an immediate and aggressive response by law enforcement, the courts and corrections. However, sex offenses are not the only crimes to receive this type of attention. The federal War on Drugs was experienced at the state and local level as well. Special task forces and interdiction programs resulted in vast numbers of arrests. At the same time, the crack epidemic hooked thousands of individuals. It is difficult to disentangle the effects of law enforcement and prosecution efforts from addiction trends. In the case of sex offenses, the trend in reduced rates of offending preceded Megan’s Law. The challenge of this analysis is to separate the effects of intervention from the existing rate reduction momentum. The first set of analyses addressed this point. The second concern is to control for historical effects. Drug offenses, like sex offenses, should reveal rate patterns consistent with intervention efforts. Other crimes should be more resistant to these specialized influences, but sensitive to larger social and political influences. The following analyses contrast the statewide sex offense trends with drug and other non-sex/non-drug offense trends.

Figure 4 displays the rates of non-sex (non-drug) offenses. The average number of crimes per 1,000 population is 50.0 with the highest rate of offending at 56 in 1989 and the lowest at 45 in 2003. As illustrated, there is a consistent increase in crime rates in the late 1980’s, followed by a five- year decline. Over the next several years the rates increased again, only to drop to their lowest levels in recent decades.

For the last five years the rate has remained stable at about 45 crimes per 1,000. In these data, there is a significant change point in 1998 (MW-U=98.00; p=.005), indicating that the levels of crime prior to 1998 were significantly higher than those after 1998.

Unlike general crime, drug-related crimes showed very different rates by year. On average, there are 68 drug crimes per 10,000 population. This varied from a high of 89 in 1989 to a low of 52 in 1985. As can be seen in Figure 5, drug crimes spiked in 1989, then dropped precipitously. Although the rates increased again following 1993, this never again approached the 1989 rate. The most recent decline appears be to be stable at around 65 crimes per 10,000 and has not achieved the 1985-86 rates. There is no significant change point.

The general decline in sex offenses in NJ is similar to that of non-sex/non-drug crimes. However, the statewide change point for sex offenses occurred during the Megan’s Law implementation year (i.e., 1994), whereas the change in trend for non-sex crimes occurred later, in 1998. The wide year-to-year fluctuations in drug crimes in fact may reflect specific policy and practice efforts, although those efforts were not sustained. In the case of sex offenses, the statewide change occurred when it was predicted to change and has maintained its impact over time.

Phase Two of the National Institute of Justice grant used a sample of sex offenders released from New Jersey Department of Corrections facilities (either the Adult Diagnostic and Treatment Center [ADTC] or one of the general population facilities) before and after the implementation of Megan’s Law. Fifty sex offenders per year (25 from the ADTC and 25 from the general population) were randomly selected for the period covering 1990 through 2000, 11 years in total. This yielded a sample of 550 cases.

For each of these cases, extensive demographic, clinical, institutional and service use, criminal history, and crime offense characteristics information was collected. This provides an opportunity to contrast outcomes (i.e., recidivism, time to failure, and harm variables) of offenders arrested and released prior to the passing of Megan’s Law with offenders arrested and released after the legislation passed.

The following sections present offender characteristics, bivariate differences in characteristics, and pre-post group outcomes.

Table 2 displays the demographic characteristics of the sample. The sample is comprised only of males. Half of the sample is white with black and Hispanic offenders accounting for 35% and 15%, respectively. Only 0.2% of offenders classified themselves as “Other.” At release, offenders were 34 years of age (sd=12.2). Nearly half (49%) were married at one time and 66 percent had at least one child (including stepchildren). On average, each individual had 1.9 children (sd=2.1).

Table 2. Demographic Characteristics of Sex Offenders (n=550)

Variable	%	Mean (sd)
Race	50.5
% white	50.5
% black	34.8
% hispanic	14.6
% other	0.2
Average Age		34.1 (12.2)
% Ever Married	49.0
% With Children	65.9
Average Number of Children		1.9 (2.1)
Education Level
% less than high school	50.3
% high school diploma/GED	33.6
% some college or more	16.1
% Ever Employed	62.8
Employment Type
% white collar/professional	7.8
% blue collar/skilled trade	75.4
% service industry	13.2
% other	3.6

Half of the sample never completed high school. Specifically, 14 percent only achieved an eighth grade education, whereas 36 percent attended high school, but did not graduate. Twenty-five percent completed high school and 8 percent obtained a GED. Sixteen percent had some college education with 4 percent completing an Associate Degree or higher. Sixty-three percent had an employment history of a year or greater prior to committing the offense. Although most offenders reported some variety of employment history, the median years of employment was considerably low, at less than three years of past employment. Of those who had been employed, most had held unskilled or trade jobs (75%) or jobs in the service industry (13%). A notable 8 percent held white-collar or professional jobs. Offenders’ prior employment income was unable to be determined for 29% of the sample. Of those offenders reporting employment income, 25% reported an income of $20,000 or less, 5% reported an income of $21,000 to $30,000, 3% reported an income of $31,000 to $40,000, 1% reported an income of $41,000 to $50,000, and 0.5% reported an annual income of $50,000 or higher.

This section includes measures commonly associated with risk (e.g., history of abuse, familial criminal justice involvement), behavioral health problems, and past treatment experiences. Table 3 displays these measures obtained from an offender’s folder.

Most offenders were raised in either a traditional two-parent home (66%) or in a mother-only headed household (23%), and the majority of offenders did not report any history of child abuse (61%). Twenty-six percent, however, reported having experienced sexual abuse as a child. A large majority of offenders (91%) did not have any family members involved in the criminal justice system.

Only 23 percent of offenders reported some type of past mental health problem. These mental health issues included problems diagnosed in childhood (e.g., emotionally disturbed, developmental disorder) as well as more common diagnoses problems such as depression. In addition, a sizeable proportion of offenders had a drug or alcohol abuse history, with 45% reporting a prior drug abuse problem and 47% reporting a prior alcohol abuse problem.

Thirty-five percent reported having received mental health treatment in the past. Most offenders (94%) were reported as receiving some type of sex offender treatment while incarcerated. A majority of offenders (88%) also received treatment in addition to the standard, mandated treatment groups. Types of adjunct treatment offered to inmates included adult basic education classes, life/social skills groups (e.g. anger management), and drug and alcohol counseling.

Offender criminal history includes information on prior arrests. These data are presented in Table 4.

In general, the men incarcerated for a sex crime were more likely to have been engaged in previous non-sex crimes than in sex crimes per se. Sixty-five percent had a previous arrest for a non-sex crime. On average, they had been arrested 3.4 times (sd=5.77) and were arrested for the first time when they were 21.5 years old (sd=8.21). Only 27 percent had been previously arrested for a violent crime with an average of .5 prior arrests (sd=1.07). Even fewer (24%) had been arrested for a sex crime in the past, with an average number of .4 prior arrests (sd=1.02). On average, these offenders were 24.8 years old (sd=9.01) at the time of their first arrest for a sex crime. Only 6 percent had been arrested as a juvenile for a sex crime.

Table 4. Offender Criminal History

Variable	%	Mean (sd)
% with Any Prior Arrests	64.9
Average Number of Arrests		3.64 (5.77)
Average Age at First Arrest		21.5 (8.21)
% with Prior Arrests for a Violent Crime	27.3
Average Number of Arrests for Violent Crime		.50 (1.07)
% with Prior Arrests for a Sex Crime	23.5
Average Number of Arrests for a Sex Crime		.43 (1.02)
Average Age at First Arrest for a Sex Crime		24.8 (9.01)
% with a Juvenile Arrest for a Sex Crime	5.7

Table 5 displays information regarding the sex crime(s) for which the men in the sample were serving sentences. Eighty percent of offenders were serving time for child molestation (incest=21% vs. non-incest=59%). Cases of rape and general exhibitionism accounted for 20% and 0.4% of all cases, respectively.

Sixty-two percent of offenders denied committing certain acts of the instant crime, or denied the sexual offense in its entirety. Most often, offenders in this latter group denied the more egregious acts of the offense (i.e. penetration) or instances of multiple acts. According to police reports, however, a majority of offenders (55%) engaged in multiple acts over a period of time, and in 26 percent of the cases the offender had multiple victims.

The 550 offenders in the sample victimized a total of 796 individuals. That is an average of 1.45 victims (sd=1.07) per offender for the current offense alone. However, this number is skewed. In 74 percent of the cases, there was only one victim identified. Of the cases involving two or more victims, the average number of victims was 2.7 (sd=1.49). Of the victims, 79 percent were female and 30 percent were male. These percentages include the cases where both males and females were victims (8%). The mean age of victim in the index offense was 12.3 years old (sd = 9.74). Ages of victims spanned from 1 year to 87 years old; 65 percent of the victims were 12 or younger, 24 percent were between 13 years old and 18, and the remaining 11 percent were 19 or older.

Table 5. Characteristics of Target Crime

Variable	%	Mean (sd)
Offense Type
% child molestation	79.5
% rape	20.2
% exhibitionism/voyeurism	0.4
% Offender Denied Some or All Aspects of Crime	62.2
% Cases Occurring Over Multiple Dates	55.2
% Cases Involving Multiple Victims Victim Gender	26.0
% male	21.1
% female	70.5
% both	8.4
Mean Age of Victim		12.3 (9.74)
Age Group of Victims	1.4 (1.1)
% 12 and under	65.4 (11)
% 13 through 18	23.7
% 19 or older	10.9
Relationship of Offender to Victim
% stranger	16.1
% family	48.2
% acquaintance	33.6
% significant other	2.2
% Lived With Victim	42.6
% Crime Occurred in Victim or Offender Home	77.2
% Cases Involving Weapon Use	13.2
Type of Weapon
% gun	27.3
% knife	51.5
% rope/tape/bondage	7.6
% other	13.6
% Drugs Involved in Crime	13.4
% Alcohol Involved in Crime	26.0

Most offenders had an established prior relationship with their victims, with only 16 percent of cases where the perpetrator was a stranger. In fact, nearly half (48%) of the perpetrators were family members, with the remaining crimes committed by either acquaintances of victims (34%) or victims’ significant others (2%). Further, 43 percent of offenders lived with their victim(s) and in 77 percent of the cases the offense(s) were committed in the victim’s or offender’s home (including shared residence).

In 13 percent of the cases a weapon was used. Of those cases, the most common weapon used was a knife (52%), followed by a gun (27%), other weapon (14%) or the use of some form of restraint (8%). In 13 percent of the cases drugs were involved and alcohol was involved in 26 percent of the offenses.

On average, offenders were sentenced to nearly nine years of incarceration (104 months, sd= 63.8), with the most frequently imposed sentence being five years. The minimum and maximum imposed sentences for the sample were one year and 36 years, respectively. In actuality, offenders served approximately five years on average (56 months, sd=40.4), with time served ranging from three months to 21.5 years. Only 32 percent of offenders were paroled whereas 68 percent maxed out; leaving the prison with no post-incarceration supervision requirements other than those imposed by Megan’s Law.

In studies that use random sampling it is assumed that the samples will be equivalent in all relevant factors. This is, however, an assumption, and statistical theory suggests that although rare, samples may be found to differ. In this case, it is known that samples differ temporally. The differences in cohorts may be reflected in institutional responses (e.g., changes in court procedures. In this case “Truth in Sentencing” legislation came into effect during this period), social or community behavior (e.g., increases or drops in specific drugs of choice or type of crime), or other historical sociopolitical changes. Bivariate analyses were conducted to confirm offender similarity in: demographics, risk factors, and prior criminality; all known to be associated with the likelihood of recidivism.

No statistically significant differences were found in demographic characteristics. Among the risk factors, only receipt of other treatment services was significant (with the earlier cohort more likely to have received services [95% vs. 83%; χ2 = 14.6, df=1, p<.001]). In terms of criminal history, no variable was found to be significant except for the average number of prior sex offenses (with the earlier cohort averaging a higher number [.56, sd=1.16 vs. .32, sd=.87; F=7.21, df=1, 546, p=.007]). Among the target offense variables, only alcohol use was significant (with the earlier cohort more likely to have used alcohol during the commission of the crime [31% vs. 22%; χ2 = 6.09, df=1, p=.014]). Thus of the over fifty variables analyzed, only three were significantly different between groups. Again appealing to statistical theory, with multiple tests there is an increased likelihood of detecting significant relationships. No correction was made in these analyses to account for this threat. However, given the vast number of equivalencies, these groups are assumed equal for purposes of the outcome analyses.

Before presenting pre-post contrasts that are controlled by time at risk, year-by-year graphs demonstrate several important points that must be kept in mind when interpreting the remainder of the analyses. The outcome measure of recidivism was collected through June 15, 2007. The remaining measures were adjusted to assure that all offenders had an equal time at risk, specifically 2,358 days or approximately six and a half years.

Figure 6 presents the percent of offenders released in each year who generally recidivate within the follow-up period (i.e., 6 1/2 years). In this case, this figure presents the percent of persons who are re-arrested, the percent of the sample re-convicted and the percent re-incarcerated. Clearly, these are three closely linked outcome measures (e.g., conviction cannot occur in the absence of a chargeable offense).

Overall, 46 percent of offenders were re-arrested (9 percent were re-arrested for a sex crime), 41 percent were convicted, and 35 percent were re-incarcerated. Although the figure shows substantial movement up and down over time, there are no significant differences by year (this is largely a power problem). Further, excluding the year 1995, all measures of recidivism are declining over time from highs in the 50 to 60 percent range in the 1990 release cohort to the 25 to 40 percent range in the 2000 release cohort. What is interesting about this figure, however, is the rates relative to each other within year. In most years, a stable percentage of persons who are arrested are convicted. In this sample, over the 11 years, 88 percent who are arrested are convicted. Of those convicted, 86 percent are incarcerated as a result. However, these rates vary from year to year. For example, of the 1993 release cohort 46 percent were re-arrested; of those, 96 percent were convicted; and of those convicted 96 percent went back to prison. In comparison, of the 1995 release cohort, 56 percent were re-arrested and nearly all were convicted (96%), but only 70 percent of those convicted were re-incarcerated. It is not clear from these data whether the year-to-year differences are a result of procedural and administrative changes or a reflection of a system response to public pressure.

Table 7 presents the comparisons of the pre and post-implementation groups on all outcome measures, including recidivism, community tenure and harm (sexual re-offending). In the first “recidivism” section, all measures (i.e., arrest, conviction and incarceration) are significant. In all three variables, the post-implementation group has a lower percentage of cases that have experienced the outcome. This is for general recidivism. Forty-one percent of the post-implementation group was re-arrested compared to 50 percent of the pre-implementation group (χ2= 3.94, 1 df, p=.047). Similarly, 34 percent of the post-implementation group was convicted compared to 46 percent of the pre-implementation group (χ2= 8.59, 1 df, p=.003). And 29 percent of the post-implementation group was re-incarcerated compared to 40 percent of the pre-implementation group (χ2= 7.53, 1 df, p=.006).

Table 7. Offender Outcomes Pre and Post Megan’s Law Implementation (n=550)

Variable	Pre	Post	Total	X²/F (df)	sig.
Recidivism
% re-arrested any crime	49.7	41.2	45.8	3.94 (1)	.047
% re-convicted at least once	46.3	34.0	40.7	8.59 (1)	.003
% re-incarcerated at least once	40.0	28.8	34.9	7.53 (1)	.006
Community Tenure
Days to arrest any crime (sd)	772.2(636.9)	726.0(616.5)	753.3(627.8)	.329(1,250)	n.s.
Days to arrest sex crime (sd)	813.7(690.5)	765.3(706.0)	794.9(689.6)	.056(1,47)	n.s.
Harm
% re-arrested sex crime	10.0	7.6	8.9	.97 (1)	n.s.
Sex crime type (n=48)				1.70 (2)	n.s.
% child molestation	54.5	66.7	59.5
% rape	13.6	20.0	16.2
% other (voyeurism, exhibitionism)	31.8	13.3	24.3
% violent	31.9	20.5	26.7	9.01 (1)	.003

Time to failure is an important outcome measure. Situations may exist where equal percentages of experimental and comparison groups demonstrate an outcome, in this case, re-arrest, but the average length of time to the arrest differs. Even in the case where equal percentages of pre and post-implementation subjects are re-arrested, more days in the community without committing a crime^[2] reflects improved outcomes in community and personal harm, as well as cost savings.

The average time to an arrest for any type of crime was 753 days (sd=628) or about two years, one month (see Table 6). There was no significant difference by implementation cohort. The average time to an arrest for a sex offense was 795 days (sd=690) or about two years, two months. There was no significant difference by implementation cohort for this variable.

A survival analysis was also conducted on these data to determine whether the rate of failure by time at risk varies significantly by implementation cohort. Figure 7 displays the survival curves for the two groups. Cases that experienced an arrest are designated by their inclusion in the continuous curve (i.e., continuous line), cases that were not arrested are censored and are represented as pluses. The strength of this analysis is the inclusion of censored cases. They are included with the time value computed as the time from release until the last day of data collection (i.e., June 15, 2007).

The curves reflect several facts: (1) all cases are censored if their time at risk exceeds 2358 days regardless of whether they were arrested or not, (2) 60 percent of post-implementation cases compared to 50 percent of pre-implementation cases survive (i.e., have not been arrested)^[3], and therefore visually demonstrating the cohort difference in overall re-offending, and (3) the curves, while diverging a small amount, are proportionally similar across time at risk, thus reflecting no significant difference in the failure rate (confirmed by statistical tests, including the log-rank test).

Re-arrests for sexual offenses do not significantly differ year to year (see Figure 8).

Holding time at risk constant, 9 percent of the sample has been re-arrested for a sex crime, representing about 19 percent of the arrest charges. This varies from a high of 14 percent in 1991 and 1992 to a low of 6 percent in 1994, 1995, and 2000.

Pre and post-implementation groups do not differ in the percent of persons re-arrested for a sex crime (10% vs. 7.6%). Of the 484 cases represented in the sexual re-offense type analysis, 60 percent were charged with child molestation or incest, 16 percent with rape and 24 percent with another type of sex offense, including voyeurism and exhibitionism. The pre and post-implementation groups also did not differ significantly on sex offense type.

As a side note, the percentage of violent crimes, excluding sex crimes, was also investigated. Overall, 28 percent of the sample was re-arrested for at least one violent crime. Importantly, only 21 percent of persons released after Megan’s Law was implemented were re-arrested for a violent crime compared to 32 percent of the pre-implementation cohort.

The final stage of this research grant proved to be the most challenging, as delineating costs associated with community registration and notification were difficult to disentangle from other state and county level spending. The research team mailed a cost assessment questionnaire to the Megan’s Law Units housed within each of the 21 county prosecutor’s office. Megan’s Law Units are responsible for the enforcement and administration of community notification and registration statutes in New Jersey (i.e., Megan’s Law). Examples of functions performed by Megan’s Law Unit personnel include risk assessment (i.e., tier classification), door to door/community notifications, trainings (e.g., law enforcement, day care center employees), prosecution/litigation, internet registry maintenance, etc.

Prior to mailing the cost assessment questionnaires, the research team met with Assistant Prosecutors in order to review questions contained in the survey and to address any questions prosecutors may have had in completing the survey. Survey questions were subsumed under two general categories: start up costs and ongoing yearly implementation costs.

Specifically, startup costs include those initial costs associated with the establishment of each county’s Megan’s Law Unit. Three variables were included under startup costs: establishment of the internet sex offender registry, equipment costs, and other/miscellaneous costs (e.g. computer software). Ongoing costs consist of expenses such as staff salaries, internet registry maintenance, equipment maintenance/supplies, and other/miscellaneous expenses (e.g. mailings, printings, software updates, etc.). Survey questions concerning on-going expenses pertained to costs accumulated during the calendar year ending 2006.

In addition, a section concerning percentage of time allotted to job tasks (i.e. itemized according to staff title) was included and was to be completed for all staff working within each county’s Megan’s Law unit. For example, if an investigator was included under personnel, a percentage breakdown of time allotted to specific job functions such as risk assessment, door to door notifications, training, etc. was required.

Of the 21 counties that were surveyed, 15 surveys were completed and received by the research unit, for a total response rate of 71.4 percent. Upon receipt, researchers scanned survey responses for possible misreading/interpretation issues related to specific survey items. For additional clarification, researchers called county prosecutor offices to confirm questionable survey item responses and made any changes accordingly. After survey responses were finalized, an Excel database of cost assessment variables was created for analysis.

Along with the cost assessment survey, prior New Jersey state budgets were reviewed for costs associated with the incarceration, rehabilitation, and tracking of sex offenders. Specifically, the budgets were searched for any allocation to Megan’s Law. Moreover, original grant documentation and archived folders were also reviewed for costs not included or found in the other sources. Sources were challenging to locate, as was the origin of much of the funding. Results

The results that follow include statistics based on the 15 counties that responded to the research unit’s Megan’s Law Cost Assessment Survey. For the 15 responding counties, the initial aggregate implementation cost of Megan’s Law totaled $555,565. Of this total startup cost, establishment of the internet sex offender registry accounted for $186,190, an average of $31,032 (sd=$24,140) per county, equipment accounted for $232,407 ($19,367 average per county, sd=$14,212), and other/miscellaneous costs accounted for $136,968 ($12,452 average per county, sd= $17,702). In addition, total aggregate expenses for all 15 counties attributable to the ongoing implementation of Megan’s Law were estimated to be $3,973,932 per annum (i.e., according to the fifteen participating counties). Of total per annum costs, staffing costs accounts for $3,605,972 ($257,569 average per county, sd= $160,180), internet sex offender registry maintenance accounts for $146,300 ($20,900 average per county, sd=$20,178), equipment/supplies accounts for $130,483 ($10,037 average per county, sd= $8,196), and other/miscellaneous expenses accounts for $91,177 ($6,513 average per county, sd= $6,002).

Additional information gathered from the prosecutor’s surveys includes counts of staff within each county’s Megan’s Law Unit, number of cases handled per year, and number of door to door notifications per year. According to completed surveys, the number of employees dedicated to Megan’s Law Unit operations totals 78 (5.2 average per county, sd= 3.2), and an estimated 5,873 Megan’s Law specific cases were processed (391.5 average per county, sd= 303.4). Moreover, counties reported that law enforcement officers performed a total of 31 door-to-door notification events (3.9 average per county, sd= 2.7) throughout the year (e.g. 1 event equals 300 households) for tier three sex offenders.

A question concerning ongoing costs for the calendar year ending 2006 was also included in the survey to measure yearly cost increases/decreases. The cost for Megan’s Law implementation during calendar year 2006 was estimated to be $1,557,978, whereas implementation costs during calendar year 2007 totaled $3,973,932 for responding counties^[4]. This change represents a 155% increase in ongoing expenses from calendar year 2006 to calendar year 2007.

These increases were obtained from raw figures provided by the Megan’s Law Units and did not reflect specific costs. However, with the inception of the Global Positioning Satellites used for Tier 3 sex offenders, it can be surmised that a portion of the increases can be attributed to increased surveillance. Finally, research of prior state budgets documented a $200,000 expenditure on Megan’s Law DNA Testing for fiscal years since 2000. There are no other distinguishable appropriations. Most costs are combined with salaries or another type of operating expenses.

The three phases of this study were designed to test the effectiveness and cost of Megan’s Law using multiple methods and strategies. In none of the analyses was Megan’s Law definitively found to be effective. Since sex crime rates have been down prior to Megan’s Law and pre and post samples do not indicate statistically lower rates of sexual offending, the high costs associated with Megan’s Law are called into question.

As a preliminary step in assessing the effect of community registration and notification laws on sexual arrest rates in New Jersey, the goal of the trend study was to explore crime trends and to identify possible changes over a 21-year period. Specifically, the main research areas concerned the patterns of sexual offense rates both prior and subsequent to the implementation of Megan’s Law, as well as comparisons in crime rates between sexual, drug, and non-sex/non-drug based offenses during the same time period. The results presented in this report support findings by other researchers exploring relevant topics. Most notably, Finkelhor and Jones (2004) found that there has been a consistent downward trend in child sexual abuses cases since the early 1990s.

This trend analysis did indeed find a significant change in the statewide decreasing sex offense rate in the year Megan’s Law was implemented, which may lead some readers to believe that the legislation is solely responsible for the decline. Because sex offense rates began to decline well before the passage of Megan’s Law, the legislation itself cannot be the cause of the drop in general. It may, in fact, be the case that continuing reductions in sex offending in New Jersey, as well as across the nation, are a reflection of greater societal changes. Having said this, it is nevertheless hard to explain the steeper decline in rates after the implementation of Megan’s Law. Given that sex offenses are low base rate events, the finding that these rates continue to decline at an accelerated rate after 1994 suggests that something other than a natural decline may be responsible. Although the initial decline cannot be attributed to Megan’s Law, the continued decline may, in fact, be related in some way to registration and notification activities. However, there may well be additional factors causing this steeper rate of decline after 1994, perhaps some attributable to other public policies. For example, in 1998, New Jersey began civilly committing those sex offenders found to present the highest risk to the community, termed sexually violent predators. Assuming the accuracy of the risk assessment that underlies the civil commitment of these sexually violent predators, then those at highest risk to reoffend have been removed from the community, thereby potentially lowering the sex offense rate, although, the number of civilly committed sexual predators only includes approximately 350 sex offenders.

Moreover, this statewide finding of a declining sex offense rate should be taken with considerable caution. The variation in the pre-post-implementation rate trends at the county level suggests that the statewide effect may be an artifact of the aggregation process. Although many counties (9 of 21) follow the state trend, many others show no differences in rates over time or have experienced reductions followed by increases to near pre-Megan’s Law levels.

Even so, with only two exceptions, the rates of sex offending were highest prior to 1994 and lowest after 1995, with the most recent years having the lowest rates. Differences in population, socio-political status, policing and prosecutorial resources may be related to differences in the effectiveness in notification and surveillance activities in specific counties.

Although impossible to distinguish the nature of the effects, the reductions of sex offenses is related to some historical process: either (1) registry/notification, surveillance and/or aggressive prosecution under a more mature Megan’s Law is responsible for the continued reductions or (2) general public awareness, publicity, and/or exclusion and intolerance feed the continued decline. Most likely, it is a combination of these factors.

In the offender release sample, there is a consistent downward trend in re-arrests, reconvictions and re-incarcerations over time similar to that observed in the trend study, except in 1995, when all measures spiked to a high for that period. This resulted in significant differences between cohorts (i.e., those released prior to and after Megan’s Law was implemented). Similarly, re-arrests for violent crime (whether sexual or not) also declined steadily over the same period resulting in a significant difference between cohorts (i.e., those released prior to and after Megan’s Law was implemented). However, because these trends began before Megan’s Law was passed, this decline cannot be attributed solely to Megan’s Law activities.

In all other pre-post measures, including other measures of recidivism, community tenure and harm reduction (decreased sexual offending), no significant differences between cohorts were found. As such, Megan’s Law does not illustrate effectiveness in:

Costs associated with the initial implementation of Megan’s Law, as well as ongoing expenditures, continue to grow over time. Start up costs totaled $555,565 in 1994 and now current costs (in 2007) total approximately 3.9 million dollars. Given the lack of demonstrated effect of Megan’s Law, the researchers are hard-pressed to determine that the escalating costs are justifiable.

Conducting a study of this type with sensitive sexual arrest data introduces a number of limitations. The most noted problem plaguing sexual offense research, the low base rate of reported sexual offenses, is tied to the under-representation of official data. Because sexual offenses are under-reported, most measures of recidivism under-represent the true offending rates (American Psychiatric Association [APA], 1999; Belknap, 2000; Furby et al., 1989; Hall, 1995; Hanson & Bussiere, 1998). It has been suggested that the present statistics on sexual abuse represent approximately one-third of the number of actual victimizations, leaving researchers and practitioners concerned about the “dark figure” of sexual abuse (APA, 1999; Belknap, 2000; Chesney- Lind, 1997). Legal definitions, fear and shame, and a desire for privacy are the main contributors to the unwillingness of many victims to report their abuse. Conversely, it has been noted that some types of sexual abuse may be over-represented to the police, such as stranger rapes (Belknap, 2000).

For example, victims of stranger rape, as opposed to incest victims, may be more inclined to report their sexual victimization because their perpetrator is unknown. This disparity may lead many to believe that stranger victimizations occur more frequently than other types of sexual victimizations because the reports may appear disproportionately higher (Zgoba & Simon, 2004). Although most individuals know that acquaintance or familial crimes are more frequent, these factors may make it difficult to achieve a clear picture of sexual offense rates (Belknap, 2000; Chesney- Lind, 1997). Given this low base rate of reporting, it is notable that sex offenses decrease rapidly in the post-Megan’s law period; the fact that the decrease accelerates as the number of crimes decreases is unexpected.

Another issue that has been difficult to fully address in the format of this study is whether the noted decreases in the post-Megan’s law period can be attributed to specific deterrence or a more general deterrent effect. The intent of Megan’s law was to reduce repeat arrests among known sex offenders. That is, Megan’s law was designed as a specific deterrent. However, the idea of notification and increased surveillance may have a general deterrent effect. Further, increased attention and public contempt of sex offenses and offenders may also contribute to general deterrence. This study illustrated downward trends in sexual arrest rates, but cannot differentiate whether the reduction is due to decreases in new first-time sex offenses (general deterrence) or to decreases in sexual re-offenses (specific deterrence).

One of the largest challenges, and a subsequent limitation, associated with this grant was obtaining the financial costs regarding Megan’s Law. County Prosecutor Offices, as well as the offices dealing with Treasury and Budget, had the same difficulties the researchers experienced when attempting to isolate and identify the costs listed in the State of New Jersey Budgets. Furthermore, initial start-up costs were sometimes funded through grants that providing few specifics regarding disbursement patterns. In an effort to provide close estimates, the researchers developed proxy measures that should be read with some caution.

Despite wide community support for these laws, there is little evidence to date, including this study, to support a claim that Megan’s Law is effective in reducing either new first-time sex offenses or sexual re-offenses. Continuing research should focus on matching samples of sex offenders before and after the implementation of Megan’s Law and also examining levels of supervision associated with Megan’s Law. Further research will be conducted utilizing the data accumulated here, specifically exploring low base rate offending and potential predictors of sexual recidivism. Should future studies establish that Megan’s Law has no demonstrable effect on the rates of sexual offending, policy makers and legislative leaders should investigate other options for lowering sex offense rates, such as mandated treatment of all sex offenders, potential use of polygraph testing and intensive probation and parole supervision.

· American Psychiatric Association. (1999). Dangerous sexual offenders: A task force report of the American Psychiatric Association. Washington, DC.

· Barnoski, R. (2006). Sex Offender Sentencing in Washington State: Sex Offender Risk Level Classification Tool and Recidivism. Washington State Institute for Public Policy, (WSIPP Publication No. 06-01-1204). Retrieved February 3, 2007 from http://www.wsipp.wa.gov/pub.asp?docid=06-01-1207.

· Beck, V. S. & Travis, L. F. III. (2002). Sex offender notification and protective behavior. Paper

· presented at the 54th annual meeting of the American Society of Criminology, Boston, MA.

· Beck, V. S., Clingermayer, J., Ramsey, R. J., & Travis, L. F. III. (2004). Community responses to sex offenders. Journal of Psychiatry & Law, 32, 141-168.

· Belknap, J. (2000). Invisible women: Gender, crime and justice. Stamford, CT: Wadsworth Publishing.

· Brooks, A. (1996). Megan’s Law: Constitutionality and Policy. Criminal Justice Ethics, 15 (1), 56-66.

· Chesney-Lind, M. (1997). The Female offender: Girls, women and crime. Thousand Oaks, CA: Sage Publications.

· Corrigan, R. (2006). Making Meaning of Megan’s Law. Law & Social Inquiry, 31(2), 267-312.

· Finkelhor, D., & Jones, L. M. (2004). Explanations for the decline in child sexual abuse cases. Washington, D.C.: Office of Juvenile Justice and Delinquency Prevention.

· Furby, L., Weinrott, M. & Blackshaw, L. (1989). Sexual offender recidivism: A review. Psychological Bulletin, 105, 3-30.

· Hall, G. (1995). Sexual offender recidivism revisited: A meta-analysis of recent treatment studies. Journal of Consulting and Clinical Psychology, 63 (5), 802-809.

· Hanson, R. K. & Bussiere, M.T. (1998). Predicting Relapse: A Meta-Analysis of Sexual

· Offender Recidivism Studies. Journal of Consulting and Clinical Psychology, 66(2), 348¬ 362.

· Langan, P.A. & Levin, D.J. (2002). Recidivism of Prisoners Released in 1994. U.S. Department of Justice, Bureau of Justice Statistics (NCJ 193427. Washington, DC.

· Matson, S. & Lieb, R. (1997, October). Megan’s Law: A review of state and federal legislation. Washington State Institute for Public Policy (Document No. 97-10-1101). Olympia, WA.

· Pettitt, AN. (1979). A non-parametric approach to the change-point problem. Applied Statistics 28:126-35.

· Pawson, R. D. (2002). Does Megan’s Law work?: A theory-driven systematic review. Centre for Evidence Based Policy and Practice, London, UK: University of London.

· Presser, L. & Gunnison, E. (1999). Strange Bedfellows: Is Sex Offender Notification a Form of Community Justice? Crime & Delinquency, 45(3), 299-315.

· Pallone, N.J., Hennessy, J.J. & Voelbel, G.T. (1998). Identifying Pedophiles “Eligible” for Community Notification Under Megan’s Law: A Multivariate Model for Actuarially Anchored Decisions. Journal of Offender Rehabilitation, 28(1/2), 41-60.

· Rudin, J. (1996). Megan’s law: Can it stop sexual predators, and at what cost to constitutionality? Criminal Justice, 11(3), 2-63.

· Schram, D. D., and Milloy, C. D. (1995). Community Notification: A Study of Offender Characteristics and Recidivism. Washington State Institute for Public Policy. Seattle, Washington: Urban Policy Research.

· Tewksbury, R. (2005). Collateral Consequences of Sex Offender Registration. Journal of Contemporary Criminal Justice, 21(1), 67-81.

· Tewksbury, R. & Lees, M. (2006). Perceptions of Sex Offender Registration: Collateral Consequences and Community Experiences. Sociological Spectrum, 26(3), 309-334.

· Witt, P.H. & Barone, N.M. (2004). Assessing sex offender risk: New Jersey’s methods. Federal Sentencing Reporter, 16, 170-175.

· Zevitz, R. G. & Farkas, M. A. (2000). Sex offender community notification: Assessing the impact in Wisconsin. U. S. Department of Justice, National Institute of Justice, Research in Brief (NCJ 17992).

· Zgoba, K. & Simon, L.M.J. (2005). Recidivism Rates of Sex Offenders Up to Seven Years Later: Does Treatment Matter? Criminal Justice Review, 30(2), 155-173.

[1] No formal statistical tests were performed contrasting drug and non-sexual offense rates to sex offense rates. The visual displays are used to provide general contrasts, only.

[2] Assuming that supervision and surveillance practices are equivalent and that the individual does not indeed commit any crimes.

[3] These are the same percentages in reverse (100%-%arrested) as those displayed in Table 1

[4] As noted, fifteen out of a possible twenty one counties in New Jersey responded to the cost assessment survey, which translates to an approximate response rate of 71.4%. In order to provide a more accurate assessment of initial and on going costs statewide, said costs were interpolated by adding 28.6% (i.e. 100% - 71.4%) to the implementation and ongoing grand totals of the fifteen responding counties. . In effect, using this general interpolation method, implementation costs were estimated to be $714,457 and ongoing costs were estimated to be $5,110,477 statewide. Again, because these figures are interpolated, these costs do not take varying county demographics into account and should be interpreted with this caveat.